By: Sam Koslowsky, Senior Analytic Consultant, Harte HanksYoure finally upgrading to a top of the line smartphone. It has the features you desire, and youre ready to pay for it. Wait a minute, the sales associate states. You might want to consider an extended warranty. Good idea, you agree. And, of course, there are those new designer phone holsters you may also want to add. Good idea, you again approve. And, of course, theres the new powerful car charger you have to have. Ok, you say. I think that makes a lot of sense. These typical everyday shopping experiences can provide the marketer with essential information. They

To view this content OR subscribe for free

Already receive the Machine Learning Times emails?The Machine Learning Times now requires legacy email subscribers to upgrade their subscription - one time only - in order to attain a password-protected login and gain complete access.

Sign up for the Newsletter with your Choice of social media account:

Read more:
What's In Your Basket? - Machine Learning Times - machine learning & data science news - The Predictive Analytics Times

GPUs became the hardware of choice for deep learning largely by coincidence. The chips were initially designed to quickly render graphics in applications such as video games. Unlike CPUs, which have four to eight complex cores for doing a variety of computation, GPUs have hundreds of simple cores that can perform only specific operationsbut the cores can tackle their operations at the same time rather than one after another, shrinking the time it takes to complete an intensive computation.

It didnt take long for the AI research community to realize that this massive parallelization also makes GPUs great for deep learning. Like graphics-rendering, deep learning involves simple mathematical calculations performed hundreds of thousands of times. In 2011, in a collaboration with chipmaker Nvidia, Google found that a computer vision model it had trained on 2,000 CPUs to distinguish cats from people could achieve the same performance when trained on only 12 GPUs. GPUs became the de facto chip for model training and inferencingthe computational process that happens when a trained model is used for the tasks it was trained for.

But GPUs also arent perfect for deep learning. For one thing, they cannot function as a standalone chip. Because they are limited in the types of operations they can perform, they must be attached to CPUs for handling everything else. GPUs also have a limited amount of cache memory, the date storage area nearest a chips processors. This means the bulk of the data is stored off-chip and must be retrieved when it is time for processing. The back-and-forth data flow ends up being a bottleneck for computation, capping the speed at which GPUs can run deep-learning algorithms.

NEURAL MAGIC

In recent years, dozens of companies have cropped up to design AI chips that circumvent these problems. The trouble is, the more specialized the hardware, the more expensive it becomes.

So Neural Magic intends to buck this trend. Instead of tinkering with the hardware, the company modified the software. It redesigned deep-learning algorithms to run more efficiently on a CPU by utilizing the chips large available memory and complex cores. While the approach loses the speed achieved through a GPUs parallelization, it reportedly gains back about the same amount of time by eliminating the need to ferry data on and off the chip. The algorithms can run on CPUs at GPU speeds, the company saysbut at a fraction of the cost. It sounds like what they have done is figured out a way to take advantage of the memory of the CPU in a way that people havent before, Thompson says.

Neural Magic believes there may be a few reasons why no one took this approach previously. First, its counterintuitive. The idea that deep learning needs specialized hardware is so entrenched that other approaches may easily be overlooked. Second, applying AI in industry is still relatively new, and companies are just beginning to look for easier ways to deploy deep-learning algorithms. But whether the demand is deep enough for Neural Magic to take off is still unclear. The firm has been beta-testing its product with around 10 companiesonly a sliver of the broader AI industry.

We want to improve not just neural networks but also computing overall.

Neural Magic currently offers its technique for inferencing tasks in computer vision. Clients must still train their models on specialized hardware but can then use Neural Magics software to convert the trained model into a CPU-compatible format. One client, a big manufacturer of microscopy equipment, is now trialing this approach for adding on-device AI capabilities to its microscopes, says Shavit. Because the microscopes already come with a CPU, they wont need any additional hardware. By contrast, using a GPU-based deep-learning model would require the equipment to be bulkier and more power hungry.

Another client wants to use Neural Magic to process security camera footage. That would enable it to monitor the traffic in and out of a building using computers already available on site; otherwise it might have to send the footage to the cloud, which could introduce privacy issues, or acquire special hardware for every building it monitors.

Shavit says inferencing is also only the beginning. Neural Magic plans to expand its offerings in the future to help companies train their AI models on CPUs as well. We believe 10 to 20 years from now, CPUs will be the actual fabric for running machine-learning algorithms, he says.

Thompson isnt so sure. The economics have really changed around chip production, and that is going to lead to a lot more specialization, he says. Additionally, while Neural Magics technique gets more performance out of existing hardware, fundamental hardware advancements will still be the only way to continue driving computing forward. This sounds like a really good way to improve performance in neural networks, he says. But we want to improve not just neural networks but also computing overall.

See the original post:
The startup making deep learning possible without specialized hardware - MIT Technology Review

Global Machine Learning Chips Market 2020 by Manufacturers, Regions, Type and Application, Forecast to 2025 delivers an in-depth evaluation of the market with the help of quantitative and qualitative information on the global market. The report highlights the latest and upcoming industry trends as well as various industry statistics such as top vendors, product types, applications, market CAGR status, geographical regions/countries, and other factors that are anticipated to increase the growth rate of the worldwide market. The report throws light on key players, demand, and supply analysis as well as market share growth of the global Machine Learning Chips market. Various favorable aspects assessed in the report are segmentation, competitive topography, and market dynamics which include drivers, opportunities, and restraints.

The research report provides key details concerning production volume and price trends. The report provides a brief summary of the application spectrum as well as market share accumulated by each product and by each application in the global Machine Learning Chips market, along with production growth. Then, details of the estimated growth rate and product consumption to be accounted for by each application have been presented. It calculates and forecasts the market on the basis of various segments. Market dynamics influencing the market during the projection period 2015 to 2025 involving opportunities, risk, threats, drivers, restriction, and current/future trends are highlighted.

DOWNLOAD FREE SAMPLE REPORT: https://www.researchstore.biz/sample-request/29722

NOTE: Our report highlights the major issues and hazards that companies might come across due to the unprecedented outbreak of COVID-19.

A Study On Market Segments:

The report provides broad segments of the Machine Learning Chips market as per product, application, and region. All of the product and application segments are studied in detail in the report with respect to market share, growth potential, CAGR, and other deciding factors.

Prominent players of the market studied in this report are: Wave Computing, Taiwan Semiconductor Manufacturing, Intel Corporation, Graphcore, Qualcomm, Google Inc, Nvidia Corporation, IBM Corporation

Status and outlook for major applications/end users/usage area: Robotics Industry, Consumer Electronics, Automotive, Healthcare, Other,

Product type covered in the report: Neuromorphic Chip, Graphics Processing Unit (GPU) Chip, Flash Based Chip, Field Programmable Gate Array (FPGA) Chip, Other,

The report states import/export, consumption, and supply figures as well as price, cost, revenue, and gross margin by regions North America (United States, Canada and Mexico), Europe (Germany, France, UK, Russia and Italy), Asia-Pacific (China, Japan, Korea, India, Southeast Asia and Australia), South America (Brazil, Argentina, Colombia), Middle East and Africa (Saudi Arabia, UAE, Egypt, Nigeria and South Africa), and other regions can be added. This section also mentions the volume of production by region from 2015 to 2025.

ACCESS FULL REPORT: https://www.researchstore.biz/report/global-machine-learning-chips-market-29722

Objectives of The Study Are As Follows:

Customization of the Report:This report can be customized to meet the clients requirements. Please connect with our sales team ([emailprotected]), who will ensure that you get a report that suits your needs. You can also get in touch with our executives on +1-201-465-4211 to share your research requirements.

About Us

Researchstore.biz is a fully dedicated global market research agency providing thorough quantitative and qualitative analysis of extensive market research.Our corporate is identified by recognition and enthusiasm for what it offers, which unites its staff across the world.We are desired market researchers proving a reliable source of extensive market analysis on which readers can rely on. Our research team consist of some of the best market researchers, sector and analysis executives in the nation, because of which Researchstore.biz is considered as one of the most vigorous market research enterprises. Researchstore.biz finds perfect solutions according to the requirements of research with considerations of content and methods. Unique and out of the box technologies, techniques and solutions are implemented all through the research reports.

Contact UsMark StoneHead of Business DevelopmentPhone: +1-201-465-4211Email: [emailprotected]Web: http://www.researchstore.biz

Read the rest here:
Global Machine Learning Chips Market 2020 Analysis, Types, Applications, Forecast and COVID-19 Impact Analysis 2025 - Jewish Life News

In this special guest feature, Paul Scully, a Vice President at Grok, believes that sometimes its easier to look at what NOT to do in order to find an AIOps solution that will work for your company. Read on to learn more about what to avoid when it comes to finding an AIOps platform that will benefit your company. With 20 years of deep expertise in helping IT Organizations improve the reliability and efficiency of their infrastructure, Grok is intently focused on building the industrys most innovative platform to bring the best of Machine Learning to IT Operations Management.

As data grows, so, too, does the AIOps market. Forrester reports 68 percent of companies surveyed have plans to invest in AIOps-enabled monitoring solutions over the next 12 months. And Gartner estimates the size of the AIOps platform market at between $300 million and $500 million per year. It poses the question if you are going to spend millions on AIOps platforms and integrate them into your critical systems, how do you know what to look for?

Sometimes its easier to look at what NOT to do in order to find a solution that will work for your company. Read on to learn more about what to avoid when it comes to finding an AIOps platform that will benefit your company.

AVOID: Significant Retooling of Your Current Platforms

If you are looking for significant short-term benefits from an AIOps platform for IT Operations Management (ITOM), you should be wary of solutions that require replacing large portions of your current systems. Organizations that take a throw the baby out with the bathwater approach to implementing AIOps find themselves bogged down with too much to do because these projects focus on replacing much of the existing toolset. In reality, this approach only increases the complexity, cost, and timing of deploying machine learning in IT Operations.

Most ITOM systems have evolved over many years, with significant effort already invested to ingest and format data, and then integrate the data with other systems. Similarly, the work queues have also evolved to incorporate a deep knowledge of the event handling process and/or incident management process. Replacing these functions only complicates the adoption of AIOps. You should consider AIOps platforms that can easily integrate into your existing monitoring infrastructure, adding an intelligence layer to the existing footprint. This allows for a much faster deployment time as well as focusing the effort and work on what really matters: results.

AVOID: Locking into a Single ITOM Reference Architecture

There are many AIOPs platforms on the market that are extensions of existing product portfolios. These solutions typically only have good integrations with tools inside their portfolio but tend to discourage integrating outside of the ecosystem if that means replacing one of their existing solutions. This makes it difficult to replace these systems or augment them with best of breed point solutions.

When evaluating an AIOps solution customers should consider solutions that are not beholden to a single vendors ecosystem. A solution that is truly agnostic provides much more flexibility and reduced total cost of ownership over time. Think twice about AIOps platforms that:

AVOID: Approaches That Require Frequent Re-Training

Different AIOPs platforms have different requirements. Different requirements mean your teams have to be trained certain ways. Understanding the objectives of the AIOps platform is important up front since they define what the data focuses on and how the operations team will work with them.

For instance, AIOps platforms that are focused on Service Assurance need to be real-time, are required to scale and must respond in seconds. This type of solution is deployed in an environment where resources are already stretched thin, meaning teams do not have the skill set to conduct constant care and feeding of the platform (nor do they have the time to frequently retrain the algorithms). Make sure youre looking for an offering that does not require constant manual retraining and that can easily integrate different data feeds.

AVOID: Offerings with a Singular Focus

Many AIOps offerings actually only focused on a single area of artificial intelligence and ingest a single data type. For example, there are countless offerings that are focused on applying machine learning to log data while others are focused on time series data and others events. To be a complete AIOps solution for Service Assurance requires the ability to ingest Logs, Events and Performance Metrics all of them, not just one. Also, remember that this ingestion of data needs to be done against real-time, streamingdata not only historical data.

AVOID: Marketing Messages as Cover for Lack of AI

The term AI has a very broad definition, whereas the term Machine Learning is more focused, and Deep Learning even more so. However, these terms have somehow become interchangeable. They are not. Unfortunately, some vendors have capitalized on the AI boom by adding AI to their marketing messages or by adding a very small amount of AI functionality to their existing offering so they can claim their solutions is an AIOps platform. This is misleading at best and deceptive at worst.

One way you can spot deceptive marketing messages is if the offering requires a lot of manual rules. True Machine Learning solutions should not require a long list of rules be built and maintained to implement the solution. Furthermore, pay attention to the types of machine learning algorithms that are deployed in the solution. If there is only one type of algorithm such as limited anomaly detection then chances are the solution has added a minimal amount of AI capability in an attempt to put marketing ahead of technology capabilities.

AVOID: Platforms that Dont Adequately Scale

Scalability is important, especially for AI systems that have strict time constraints. AIOps systems that run against primary historical data for the purpose analytics tend have less constraints on response times from the machine learning models. However, if the system is focused on real-time data such as in a Service Assurance environment response time becomes very important.

As the business grows so does the data within the organization. When new customers are brought on, they come with new data and potentially new equipment. As new services are rolled out new data is generated and all of this new data must be captured in the AI platform. Sizing the AI platform at the beginning for the data set that exists at the time can quickly result in the system running out of resources causing response time degradation or worse system failure.

Deploying AI within a microservices architecture allows for components to more easily scale on demand. In addition, it allows components to be decentralized and scaled at the component layer versus across all components.

Knowing what to avoid when implementing AIOps is just as important as knowing what to look for. At the end of the day you want a robust platform that operates with various types of data, that does not require significant retooling of your architecture or continual retaining the algorithms, and that can scale as your data increases. Keep focused on the objectives you want to accomplish with an AIOps platform and insist on real technology, not marketing messages or limited add-ons. These real solutions exist and, once implemented, can make a considerable contribution to your Operations team.

Sign up for the free insideBIGDATAnewsletter.

View post:
AIOps 6 Things to Avoid When Selecting a Solution - insideBIGDATA